LINAGORA

LINAGORA is a leading French SME providing Open Source Software edition and Professional Services. With more than 160 employees, including computer science engineers, R&D engineers and PhD students, LINAGORA has offices in France (Paris, Toulouse, Lyon and Marseille), Canada (Montréal), Viet-Nam (Hanoi) and Tunisia (Tunis).

#AI #MachineLearning #DeepLearning #AutomaticSpeechRecognition #NLP #OpenSource #CloudComputing #Privacy

LINAGORA is currently developing the following open-source software:

• LinTO is a privacy-aware voice assistant designed to help manage business meetings (http://linto.ai). Leveraging advanced Artificial Intelligence/Machine Learning technology for speech processing and Natural Language Understanding, LinTO offers specialized features such as meeting room management and the automatic generation of meeting minutes. LinTO is available as an open-hardware device that embeds a Raspberry PI, a microphone array and a 360° camera to manage voice and face recognition.

• OpenPaaS combines Enterprise Social Network and “Platform as a Service” technologies in an advanced social platform designed for the Cloud (http://www.open-paas.org/). It offers professional and team-productivity tools such as e-mail, document editing, collaborative calendars, chat, and Web-conference tools. Advanced AI Natural Language Processing algorithms enable it to provide smart reply and contextual recommendations as well.

LINAGORA’s main research domains include:

• Automatic Speech Recognition: We are currently refining our own Automatic Speech Recognition engine using approaches from deep learning, with a particular focus on Large Vocabulary speech recognition in the challenging domain of spontaneous, multi-party conversation—a defining feature of business meetings.

• Natural Language Processing: To tackle central but difficult topics in natural language understanding, such as real-time recommendation and automatic meeting minutes, we are pioneering a new weakly supervised approach to NLP. Our model, based on the Snorkel data programming paradigm developed at Stanford, combines insights from both hybrid symbolic and deep learning methodologies.

• Cloud/Edge Computing & IoT: We are working on developing a decentralized machine learning platform based on the Cloud Computing at the Edge paradigm. Using insights from federated learning, our goal is to be able to exploit data from distinct locations simultaneously without moving that data to the Cloud, in line with the privacy-by-design ethic central to Linagora’s research.

• Privacy and Security: LINAGORA is fully committed to investigate methods and tools that facilitate the development of secure and privacy aware collaborative AI applications such as privacy preserving multi-party computation, homomorphic encryption and federated learning.

LINAGORA participates in numerous collaborative projects, funded by both French (PIA, ANR) and European (H2020, FP7) agencies.

LINAGORA’s research lab has produced several technical and academic publications based on their prototypes and research achievements (see http://research.linagora.com).

Type

Small and Medium sized Enterprise

Country

France

Website

http://www.linagora.com